Short, Medium, and Long Term Consequences of Poor Infant Health:
An Analysis using Siblings and Twins
March 2006
Phil Oreopoulos
University of Toronto and NBER
Mark Stabile
University of Toronto and NBER
Randy Walld
University of Manitoba
Leslie Roos
University of Manitoba
We gratefully acknowledge those individuals who helped make this research possible.
These included: from the Department of Education, Citizenship and Youth --John Van
Wallenghem, Richard Perrault, Carol Crera, Jean Britton, Ken Clark, and Shirley
McLellan; from the Ministry of Family Services and Housing --Harvey Stevens, Grant
Doak, Gudrun Fritz, and Jan Forster; and from the Ministry of Health -- Louis Barre. And
Janet Currie provided very helpful comments. Stabile thanks the Canadian Institutes for
Health Research for financial support. We also thank Florian Hoffman for excellent
research assistance.
ABSTRACT
We use administrative data on a sample of births between 1978 and 1985 to investigate
the short, medium and long-term consequences of poor infant health. Our findings offer
several advances to the existing literature on the effects of early infant health on
subsequent health, education, and labor force attachment. First, we use a large sample of
both siblings and twins, second we use a variety of measures of infant health, and finally,
we track children through their schooling years and into the labor force. Our findings
suggest that poor infant health predicts both mortality within 1-year, and mortality up to
age 17. We also find that infant health is a strong predictor of educational and labor force
outcomes. In particular, infant health is found to predict both high school completion and
social assistance (welfare) take-up and length.
JEL Classifications: I12, I18, J13
2
1. INTRODUCTION
Infants born in poor health, as measured by low and very low birth weights and
low Apgar scores, have lower chances of survival, and may also experience further health
and social difficulties later in life (Conley, 2003). Low birth weight babies are also
increasingly expensive to treat in hospital. Almond et al. (2005) calculate that among
babies weighing 2000 grams, an additional 450 grams is associated with a $10,000
savings in hospital charges for inpatient services. As such, understanding the causes and
consequences of poor infant health has been a primary concern of both the medical and
health policy literature for some time.
Medical advice to expecting mothers on how to prevent low birth weight,
including refraining from smoking and seeking prenatal care, is centered around the
notion that preventing low birth weight will improve both the life chances of the child
and chances of future success. Researchers have also noted the potential to reduce
hospital costs significantly through inexpensive prenatal interventions aimed at reducing
low birth weight in particular (Almond et al., 2005). Program evaluations on Medicaid
expansions in the U.S. (Currie and Gruber, 1996) and the institution of national health
insurance in Canada (Hanratty, 1996) have also focused on improved prenatal treatment
and its potential effects on infant health, providing further evidence of the policy
importance of, and potential benefits associated with improving infant health.
As noted in Almond et al. (2005), interventions aimed particularly at reducing low
birth weight are premised on the notion that low birth weight in particular is the cause of
poor health and related outcomes in the future, and not simply a marker and correlate of
such problems. While interventions and public policy aimed at improving overall infant
1
health, including reducing the incidence of low birth weight, are likely to have both short
and long term benefits, a clearer understanding of the causes and consequences of poor
infant health can only help to improve the efficacy of both health care and public policy.
An analysis of the long-term impact of infant health may also uncover important
relationships not realized from focusing on earlier outcomes. Infants born lower than
average birth weight but not considered at risk of early death, for example, may in fact
benefit from prenatal care. Or, the majority of low birth weight infants that survive past
one year may face few subsequent risks.A considerable body of research attempts to
quantify the effects of early infant health on both early childhood survival and on future
health, education, and social outcomes.
Conley, Strully, and Bennett (2003), for
example, examined the effects of birth weight for both fraternal and identical twins on
both neonatal and post-neonatal mortality. Using data from the Matched Multiple Birth
Data Set, 1995-1997, which contains data on 271,000 twins, they conclude that birth
weight differences between twins affects infant mortality and that this effect is stronger
for fraternal than identical twins.
Almond, Chay, and Lee (2005) examine the
relationship between low birth weight, low Apgar scores, and mortality in the first year of
life. Using a large sample of twin births from the National Center for Health Statistics,
they show that, while both birth weight and Apgar scores are strongly related to infant
mortality across families, the relationship between birth weight and infant mortality
significantly decreases when differences between twins are examined. In contrast, the
relationship between Apgar scores and infant mortality remains strong both across
families and within twin pairs.
Both papers note that, while twin samples can be
extremely helpful in eliminating unobserved heterogeneity across families, the resulting
2
sample is somewhat unique in that twins tend to be of lower weight than the average in
the singleton infant population.
A second stream of social science literature has used twin studies to examine the
longer-term effects of birth weight on health and education. Behrman and Rosenzweig
(2004) use twin data from the Minnesota Twins Registry to examine the effects of low
birth weight on the educational attainment and adult health of women. They find that
increasing birth weight increases schooling attainment by about one third of a year and
that this effect is stronger within twins than across children of different families. Conley,
Strully, and Bennett (2003) examine the effects of low birth weight on high school
graduation and placement in special education. Using the Panel Study of Income
Dynamics, Conley et al. exploit within sibling variation to examine the longer-term
consequences of low birth weight, finding that the effects of low birth weight on timely
high school graduation are more pronounced among siblings than across families. The
study does not look at other measures of infant health (Apgar and gestation) nor does it
explore the potential non-linear effects of low birth weight on infant health. et al. (2005)
use a sample of Norwegian twins to examine the long-run consequences of low birth
weight. Their evidence confirms that low birth weight is not a good predictor of infant
death within twin pairs. However, they do find long-term effects of low birth weight on
cognitive outcomes, educational outcomes, and on earnings.
Our paper tries to reconcile some of the disparate results of previous work in the
following ways: First, it uses an administrative sample of both siblings and twins to
examine the effects of infant health on mortality within one year. Comparing sibling
3
findings and twin findings allows us to overcome concerns that twins are a select sample
of the population and that inference from this sample is not, therefore, generalizable to
the broader population. Second, tracking both siblings and twins through school and into
their early experiences in the workforce provides longer-term evidence for both groups,
including educational outcomes, health care costs, and social assistance receipt. Third, a
variety of infant health measures, including birth weight, Apgar scores, and gestational
length, are used to contrast the effects of these measures on outcomes and to reconcile
and expand the findings of other research using multiple measures of infant health.
Gestational length is an important determinant of low birth weight, one which twin only
studies are unable to examine. Finally, using a sample of children with uniform access to
health insurance further corrects for any potential unobserved heterogeneity within
families across siblings that might not be captured in sibling fixed effects models and
offers an interesting comparison with a U.S. sample lacking universal coverage.
Our findings offer several advances to the existing literature on the effects of
early infant health on subsequent health, education, and labor force attachment. First, we
confirm earlier results by Almond et al., which show that the effect of infant health as
measured by birth weight less than 2500 grams largely disappears when looking at within
twin variation. The 5 minute Apgar score, and measures of very low birth weight (less
than 1500grams) are stronger predictors of infant mortality within one year than birth
weight for twin samples. However, we find that within sibling pairs Apgar, low birth
weight, and gestational age predict infant mortality within one year, even though we
continue to account for unobserved heterogeneity across families. Second, infant health is
found to predict both high school completion and social assistance (welfare) take-up and
4
length. We find evidence of longer-term consequences of infant health both across
families, within siblings, and within twin pairs, although different measures of infant
health predict outcomes differently. The results suggest strong effects of infant health on
death between ages 1 and 17, grade completion, and months on social assistance after age
18, even for ranges not considered overtly concerning (e.g. birth weights between 2500
and 3500 grams and Apgar scores of 7 or 8). Interestingly, we find weaker evidence of
the longer-term effects of infant health on either cognitive ability as measured by
language arts test scores or longer-term physician visits and costs. Overall, we conclude
that there are indeed long-term consequences of poor infant health, and that a better
understanding of these consequences can be determined by examining a variety of infant
health measures and by examining the variation both within families and within twin
pairs. The implications of these findings is that reductions in poor infant health will lead
to lower mortality, greater human capital accumulation, and lower welfare usage.
The paper proceeds as follows: Section 2 describes the data used in the analyses.
Section 3 outlines our empirical methodology. Section 4 presents our results, and section
5 concludes.
2. DATA
The data are from the Population Health Research Data Repository at the
Manitoba Centre for Health Policy (MCHP). Our main data match hospital records at
birth to other administrative records on education, physician visits, and social-assistance
take-up. Hospital records at birth were checked against Manitoba Health registry data
5
(which is coordinated with federal Vital Statistics files). The sample includes over 96
percent of all children born in Manitoba in 1978-1982 and 1984-1985 and more than 99
percent of this group remaining in the province up to June of their 18th year. The cohort
born in 1983 was not included because grade 12 provincial tests were not given in the
school year 2000/2001 (when the 1983 birth cohort would be expected to be in grade 12).
Health, educational and social assistance outcomes are tracked up to 2004.1
The birth data originate from Manitoba Health hospital records. Since 1970, the
registry attaches to every birth a family identification number (called the Registration
Number or REGNO), which links the infant to the ‘family head’, usually the father.
When an individual turns eighteen years old, he or she receives his or her own REGNO.
On marriage, a female receives the REGNO of her husband.
Both the mother’s
identification number (an encrypted PHIN or Personal Health Identification Number) and
REGNO are used to define siblings. Several checks on this algorithm as applied to the
seven years of birth cohorts (looking at missing data, the number of children designated
as having the same mother and father, and complicated blended families) have indicated
it to be highly accurate.
Two siblings with the same birth date are designated as twins. The birth records
do not allow us to distinguish between monozygotic and dizygotic twins. Based on
earlier descriptive studies [e.g. Conley et al., 2003], our twins data are likely comprised
of roughly 25 percent monozygotic pairs and 75 percent dizygotic pairs.
The postal code from the family head’s address identifies the street or building
where the family lives. The address of the family is updated about every six months. To
1
In Canada, welfare is more commonly referred to as social assistance. For consistency, we maintain this
terminology.
6
proxy for general social economic background, family income in the 2001 Census was
aggregated and averaged over Enumeration Areas, which were in turn matched to
corresponding postal code addresses in our sample.
Enumeration areas contain a
population of about 400 to 700 persons. The areas were ranked from highest to lowest
income and then grouped into five population quintiles. Mustard et al. (1999) and Roos et
al. (2005) show a substantial correlation (0.435) between our measure of persons’
neighborhood average income and self-reported household income (not available in our
data).
Table 1 presents descriptive statistics of the infant health measures recorded on
the hospital records and used in our study: birth weight (in grams), gestation (in weeks),
and 5 minute Apgar score (on a 10 point scale).
Means, standard deviations, and
percentiles for these measures are shown for the full sample of births between 1979 and
1985. These statistics are also shown for the subset sample of births with at least two
siblings identified within this cohort range and the subset sample of twins. The sibling
sample excludes twins.
The frequency distributions of these variables compare similarly with those
generated from nationally representative samples of Canada or the United States. The
mean birth weight among the full sample is about 3,500 grams. Twins weigh about 950
grams less and are born about three weeks earlier. About 7 percent of the full sample is
born low birth weight, defined as weighing less than 2,500 grams. In the analysis below,
we explore not only the effects of being born less than 2,500 grams and less than 1,500
grams, but also the effects of being born below average birth weight, between 2,500 to
3,000 grams and between 3,001 to 3,500 grams. Gestation before birth typically takes
7
about 40 weeks. Preterm births are often defined as births before 37 weeks gestation;
there are 7 percent preterm births in our full sample. Late births occur after 41 weeks.
The Apgar score summarizes 5 vital sign conditions at birth. Heath care providers
assess an infant’s heart-rate, respiration, muscle tone, reflex, and color and assign values
of 0, 1, or 2 for each category, with the best possible total score equaling 10. A score less
than seven often triggers additional action to stabilize conditions. A score of 7 to 10 is
considered normal. As shown below, lower 5 minute Apgar scores even within this
normal range affect subsequent educational outcomes and social assistance take-up.
The typical variation in these infant health measures between a pair of siblings or
a pair of twins is about 55 to 70 percent of the typical variation between any randomly
chosen infant pair. Column 2 of Table 1 lists standard deviation for each variable, across
all individuals. Column 3 shows standard deviations in these infant health measures
within families, among siblings and twins. These amounts are the standard deviations of
the residuals generated after regressing the health measures on a set of family fixed
effects. The standard deviation for Apgar scores is about 0.92 over the full sample and
0.65 within families. The standard deviation for gestation is about 2 weeks and 1 week
between siblings. The within family standard deviation of birth weight is still 314 grams
between siblings, and 202 grams between twins. In perspective, Almond, Chay, and Lee
(2005) report that the average difference in birth weight between a newborn with a
mother who smokes and one with a mother who does not is 285 grams.2 The average
difference in gestation is 0.3 weeks, while the average difference in 5 minute Apgar score
is 0.07. Our main analysis uses within family variation in infant health to explore short
2
These figures are for the sample of Pennsylvania singletons born between 1989 and 1991. The means,
standard deviation, and percentiles reported in Table 1 are similar to those reported by Almond, Chay, and
Lee for their Pennsylvania sample.
8
and long run differences in social economic outcomes. Column 3 indicates considerable
variation within families to work with in exploring later outcome differences.
We observe differences in infant health across both siblings and twins for several
reasons. Between siblings birth weight can differ due to both gestational length and
differences in intrauterine growth retardation (IUGR). Between twins differences in birth
weight are attributable to differences in IUGR. Twin studies have emphasized that much
of the literature has focused on differences in IUGR, despite the fact that gestational
length accounts for a significant percent of the low birth weight infants. One possible
reason for this, as noted by Almond et al. (2005) and reported in Goldenberg and Rouse
(1998) is that there is little medical evidence on how to effectively increase gestational
length, whereas there are widely accepted policy interventions aimed at IUGR (reducing
smoking and ensuring appropriate nutrition during pregnancy, are the most common of
these).
Apgar scores differ between both siblings and twins. The test was initially
designed to measure whether infants required immediate medical care, and has been
shown to be highly correlated with infant mortality (Almond et al., 2005) 3. After testing
whether the infant health measures presented here are good predictors of death in the first
year, we then consider, conditional on survival, whether they are also predictors of poor
health later in life and potentially measures of cognitive ability and human capital as
well.
Table 2 lists the health and socioeconomic outcomes explored in our paper. The
infant mortality variable comes from matching births and deaths from the Manitoba Vital
Statistics over the first year of life. The variable takes on the value of 1 if a birth is
3
The definition and purpose of the Apgar were obtained from the NIH web site at:
http://www.nlm.nih.gov/medlineplus/ency/article/003402.htm
9
matched to a death in the first year, and zero otherwise. A death between ages 1 and 17
is similarly recorded.
The other outcome variables came from administrative data on physician costs,
education, and social assistance. These data are available only for Manitoba residents.
The analysis of the effects of infant health on these longer-term outcomes, therefore, is
conditional on survival and conditional on remaining a resident in the province. We
focus on estimating the long-term effects of infant health for those born in Manitoba and
living in the province at least until they reach 17.5 years old.
Table 2 indicates that 24
percent of our original sample of births in Manitoba between 1979 and 1985 either died
or left the province before this age. We shall document that health at birth does indeed
affect mortality before age 17, even after one year, but it does not affect mobility. For all
outcome measures except mortality and mobility, we condition on the sample of those
remaining in Manitoba at least until age 17.
The Manitoba Repository data record hospital discharge abstracts and physician
claims extending back to 1970. Physician claims include diagnostic information and are
primarily reimbursed on a fee-for-service system. We summarize adolescent health by
summing the number of ambulatory physician visits recorded between ages 12 and 17.
An ambulatory physician visit is any contact with a physician that is billable by the
physician to Manitoba Health and occurs while the patient is not a hospital in-patient.
This includes: physician services received in hospital emergency rooms and outpatient
departments, contacts with physicians in salaried positions, consultative and nonconsultative care, and physician visits to residents of personal care homes. Excluded
from ambulatory physician visits are: all claims for optometrist, oral surgery, dental,
10
periodontal, and chiropractor contacts; inpatient visits (that is, contact with a physician
while admitted to a hospital); and laboratory tests, radiology, and similar services. Over
90 percent of this population contacts a physician over a two-year period and the average
visit rate is more than four visits per year.
Files from the Ministry of Education, Citizenship, and Youth (high school age
education data) and the Ministry of Family Services and Housing (income assistance
data) were linked to the provincial registry at Manitoba Health. After removal of all
identifying information (name, address, etc.) and attachment of an encrypted record
number, the data are sent to the Manitoba Centre for Health Policy. A crosswalk file
containing only the encrypted record number and encrypted PHIN is also provided. Only
authorized persons can access the data files and the corresponding crosswalk.
The enrollment records are used to determine whether a student has attained
Grade 12 by age 17. Not attaining grade 12 by this age could indicate a student has
dropped out or been held back in a grade at least once. This measure is available for all
seven birth cohorts used. Students may not have reached Grade 12 because they have
been held back. On the other hand, many students held back are more likely to drop out.
Our measure proxies as an overall indicator for being at risk of ending up with a low
level of education attainment.
We also include information from provincial language arts standards tests taken in
grade 12.
These tests contribute 30 percent to the students’ final course grade.
Individuals pass the language arts test by scoring 50 percent or more on a comprehensive
exam. The test focuses on reading comprehension, exploring and expanding on ideas
from texts, the management of ideas and information, and writing and editing skills. For
11
each birth cohort, we record the test score in 5 percentage point categories (13 in total,
with a residual 14th for students scoring between 0 and 35 percent) in the year that most
students write the test. Within each birth cohort, approximately 35% of test scores are
missing. For these students we impute test scores based on the reason for missing
information (ranking them below the lowest scoring category among those who wrote the
test). These additional categories, listed by highest to lowest rank are: absent (about 1
percent of each birth cohort sample); In grade 12 but not tested (about 8 percent); In
grade 11 or lower (about 19 percent), Not enrolled (about 2 percent), and Withdrawn
from School (about 10 percent). For the entire sample, we therefore have 19 test score
categories. Following methods forwarded by Mosteller and Tukey (1977) and Willms
(1986), we then compute a standardized score for each individual by assuming an
underlying logit distribution, which is divided into pieces according to the percentage of
cohort members in each category. Scores are calculated separately for each birth cohort
because of small changes in the categories available and in the percentage distribution
each year. In a typical year, the highest scorers are given an index score of 2.96, while
those withdrawn from school are given a score of -1.84. The logit transform produces an
index with an overall mean of zero and a standard deviation of one. The ordering on this
index is closely correlated with the student’s eventual graduation status (the point biserial
correlation is 0.54).
Finally, the sample of Manitoba residents is matched to monthly social assistance
records up to March 2004. Our youngest birth cohort can only be followed for about a
year after the age of 18. The oldest cohorts are followed from age 18 to age 25. Nine
percent of our sample received some social assistance before April 2004. Fixed effects
12
for birth cohort in our regressions will absorb average differences in take-up due to this
truncation of information. In case infant health may also affect the length of time on
social assistance, we focus on the number of months individuals in our sample used these
services. The average number of months on social assistance over our entire sample is
2.1.
3. EMPIRICAL METHODS
We estimate models of the effects of early infant health on mortality, health
expenditures, educational performance and social assistance receipt as follows :
yijt inf healthij X ij j t ijt
(1)
Where yijt represents the outcomes for individual i in family j, at time t . X measures
family or individual specific controls such as marital status, sex of the child, mother’s age
at birth, and, in some cases, household income measured at the zip code/ postal code
level. We also include a set of dummies for the birth order of the child within each family
size to completely control for any effects of both birth order and family size.4 The 
t are
year of birth fixed effects to account for any differences by year of birth of the child. The

j are family fixed effects which, as we outline in greater detail below, are included in
some specifications.
4
Several studies point out that family size correlates with education and other social economic outcomes.
Black, Deveraux and Salvanes, 2004, also find important differences in outcomes depending on birth order.
We control for these differences with family size and birth order fixed effects, in case these variables also
relate to infant health. Excluding such controls does not change our baseline results in significant ways.
13
Our primary parameter of interest is , which is the coefficient associated with
our estimate of the effect of early infant health. As discussed above, we use three
different measures: birth weight, 5 minute Apgar score, and gestational length in weeks.
Our main analysis classifies these infant health measures into categories and uses dummy
variables to estimate possible non-linear effects of infant health. This approach helps
uncover more detailed relationships between infant health and our outcome measures.
For example, education attainment may differ by birth weight only for the small fraction
born weighing less than 2,500 grams and surviving. In this case, a linear regression
model would not adequately capture this relationship. For Apgar score, we estimate
effects at birth by comparing scores of 6 or less, 7 to 8, or 9, to a score of 10. For birth
weight, we group infants by whether they weigh 1,000 grams or less, 1,001 to 1,500
grams, 1,501 to 2,500 grams, 2,501 grams to 3,000 grams, 3,001 to 3,500 grams, and
3,501 grams or more. For gestation, we compare normal gestation length, between 40
and 41 weeks, to infants born with less than 37 weeks gestation, 37, 38, 39 weeks of
gestation, and 42 weeks or more. To facilitate discussion and compare with some earlier
studies, results from the linear specification are presented in the Appendix.
For each measure of infant health we estimate 5 models: OLS using our entire
sample, OLS using the sample of children with siblings, OLS using the sample of twins
in the data, the sibling sample including family fixed effects (j) and finally the twin
sample including family fixed effects.
Our estimates of equation (1) serve two purposes. First, we are able to replicate
the results found in Almond et al. (2005), contrasting OLS and twin models with family
fixed effects, using a smaller sample of Canadian children. The differences using between
14
family and within family variation found in that research are shown to hold for this
sample of Canadian children as well. Second, we are able to expand on the Almond et al
analysis of the effects of infant health on one-year mortality by estimating fixed effects
models using variation in infant health across siblings instead of across twins.
In addition, we estimate alternate versions of equation (1) using the other outcome
measures described above, including: whether the child was held back a grade, the child’s
language arts test scores measured in grade 12, whether the child dropped out of high
school before graduation, and whether the child was on social assistance. Thus we are
able to apply the same OLS, twin, and sibling analyses to a variety of longer-term
measures of child health and social outcomes.
4. RESULTS
Table 3 shows the effects of our measures of infant health on infant mortality
(death before age 1). The subsequent tables presenting results with different outcome
variables have a similar structure. Column 1 displays the coefficients on the infant health
categories for the full sample of singletons and siblings, without family fixed effects.
These results are from the linear probability model for whether an infant died in the first
year regressed on the infant health dummy variables, plus controls for mother’s marital
status, gender of child, and a complete set of dummy variables for all family size and
birth order combinations. The second column shows the same regression, but for the
subset sample of births with at least one other sibling identified within the birth cohorts
15
1979 to 1985 (but excluding twins). In the third column, the coefficients presented
correspond to the regression model that now includes family fixed effects. The fourth
and fifth columns show the results among twins, without and with family fixed effects
respectively.
The first panel shows the results defining infant health by 5 minute Apgar score.
Infants born assessed with an Apgar score below 7 are about 26 percentage points more
likely to die within one year than those with Apgar scores of 10, and 31 percentage points
more likely to die among the sibling sample. This relationship holds when we use only
differences between siblings in column 3. The coefficient remains virtually unchanged.
However, after adding family fixed effects in the twin sample, the coefficient falls by
about two-thirds. The relatively higher association between Apgar and early death is far
less severe for those with scores of 7 or 8. While such assignments are not normally
considered indicators of critical need, non-twin siblings in this category are about 1.9
percentage points more likely to die within a year than other siblings with scores of 10.
For twins, however, this relationship drops by a third, and is measured less precisely
because of the smaller sample size. The results also suggest only a minute difference in
infant mortality between infants with Apgar scores of 9 versus 10.
The second panel presents the same set of results, but using birth weight instead
of Apgar categories. Interestingly, the same contrast in results between the sibling and
the twins samples arises when we compare the effects of very low levels of birth weight
on infant mortality with and without fixed effects. The estimated effects of low birth
weight slightly increase after adding the family fixed effects for the sibling sample. Even
for infants born between 2,500 and 3,500 grams – below average weight but not typically
16
considered low birth weight -- there is about a 1 percentage point higher risk of death
within one year. The estimated effect associated with weighing less than 1,500 grams
falls by about two-thirds when comparing twins from the same family compared to using
cross variation of the non-twin sibling sample. Similar results were found by Almond,
Chay, and Lee (2005), who focus on twins exclusively.
Variation in infant health between twins cannot be due to changes in socioeconomic circumstances of the parents between births. Changes in such circumstances
over the 1 to 7 year period in our sibling sample do not seem large enough to explain the
different estimates, especially since the coefficients do not fall at all after adding the
fixed effects (the twins results suggest a downward omitted variable bias). Another
explanation is that twin birth weight variation cannot result from differences in gestation,
but it certainly can with the sibling sample. The last panel indicates siblings born
premature are significantly more likely to die in one year than another sibling not born
premature. A sibling born 37 weeks since conception faces a 1.4 percentage point higher
chance of infant mortality than another sibling born between 40 and 41 weeks since
conception. We also find slightly higher chances of infant mortality from 39 weeks
gestation. Thus, one possibility to explain the different estimated effects from low birth
weight and low Apgar score between non-twin siblings and twins is gestation, an
important source of variation correlated with these measures of infant health but left out
from the between twins analysis.
Extending the analysis on child death out to the first 17 years of life we continue
to find an effect of both low Apgar scores and low birth weight on survival when we
examine children across families and between siblings within families (Table 4). Indeed,
17
the coefficients on birth weight between 1000 and 2500 grams actually increase once we
include family fixed effects, and the coefficients on the lowest Apgar scores remain
relatively stable. Using the twin sample, however, we find no evidence of a negative
relationship between infant health and death up to age 17. In fact, the coefficient on
Apgar scores between 7 and 8 is the wrong sign and marginally significant. Given the
drop in predictive power within twins for our one-year mortality rate estimates, it is
perhaps not surprising that we do not find results with twins for mortality 17 years out.
On the whole, however, we take these results as evidence that infant health continues to
be a strong predictor of mortality both across and within families even up to age 17.
Table 5 indicates no important effects from infant health on Manitoba emigration.
The OLS results with the full sample indicate a small correlation between poorer infant
health and moving away from the province before age 17. But the point estimates
gravitate towards zero after including the family fixed effects regardless of which
measure of infant health is used. Our results suggest that once we control for family
fixed effects, our estimates on the impact of infant health on later outcomes among
Manitoba residents do not appear to be biased from some fraction of our sample leaving
the province.
Conditional on survival until age 17, we find little evidence of significant effects
of infant health on physician utilization between ages 12 and 17. Table 6 displays the
estimates of the effects of Apgar score, birth weight, and gestation on total physician
visits between these ages, with and without including family fixed effects.
The
dependent variable here is number of physician visits between the ages of 12 and 17. We
find little consistent evidence here to support a relationship between infant health and
18
physician visits 12 to 17 years later. Although we do find some evidence of a greater
number of visits within twin families for those children with birth weights between 1000
and 1500 grams, the majority of the coefficient estimates are insignificant and some are
also the wrong sign.
Table 7 shows the results of the language arts standards test for the sample of
Manitoban residents at age 17. Recall that for the approximately 30 percent of residents
who did not write the test, the score is imputed by ranking these individuals lower than
those writing the test and categorizing them by enrollment and attainment categories (e.g.
withdrawn from school or held back). A score is given to each associated test score and
education attainment category using a standardized logit transformation weighted by the
population size in each group.
Columns 1 and 2 indicate a clear positive correlation between infant health and
the language arts test measure. For example, a low birth weight child averages a score
about 0.23 standard deviations below a child born weighing above 3,500 grams. Apgar
scores less than 8 and gestation lengths less than 38 weeks are also associated with
significantly lower grade 12 test scores. The relationship weakens notably after adding
the family fixed effects. The point estimates for birth weight and Apgar score are all still
negative, but many are no longer significant. Siblings given an Apgar of 6 or less receive
a test score about one-tenth of a standard deviation lower than a sibling with a 10. We
also find indications of small, but in some cases significant, lasting effects on test scores
from being born low birth weight, even among youths born weighing between 2,500 and
3,500 grams (only slightly below the average). The negative coefficients associated with
gestation of less than 37 weeks without fixed effects become close to zero once fixed
19
effects are added. All the point estimates for the gestation results are very small and
insignificant. The large standard errors around the estimated effects for the twins sample
(with or without fixed effects) make it difficult to draw any insight from the analysis
using gestational length here.
While our results for language arts test scores are mixed, our results examining
high school attainment suggest long-lasting effects of infant health. Table 8 shows the
estimates for the effects of infant health on reaching grade 12 by age 17. An individual
may fail to reach this grade because she either dropped out or repeated at least one earlier
grade. The sibling fixed effects analysis in column three indicates a substantial impact on
grade 12 attainment from infant health. Moving from the results with no family fixed
effects to those that include them show only a small fall in the coefficients.
A newborn assessed with an Apgar score of 6 or less, for example, has a 7.4
percentage point lower probability of reaching grade 12 by age 17 compared to a youth
born with an Apgar of 10 (column 2). When family fixed effects are added (column 3),
the estimated effect drops to a 4.1 percentage point difference in the probability of
reaching grade 12 by this age. Siblings with Apgar scores still considered normal but
below average (7 or 8) are 2 percentage points more likely to drop out or repeat a grade.
A similar story holds when looking at birth weight. Students born with low birth
weights between 1500 and 2500 grams, and who survive until age 17 are about 8
percentage points less likely to be enrolled in grade 12 than those born weighing 3,500
grams or more. The chances of infants attaining Grade 12 by age 17 are severely affected
by being born weighing less than 1500 grams. The gestation results are also strong and
significant. Premature siblings born in 36 weeks or less are 4.0 percentage points less
20
likely to have reached grade 12 by age 17 than those born in 40 to 41 weeks. Negative
effects on this measure of educational attainment among those born in 38 or 39 weeks are
also detected. A sibling born after 38 weeks gestation is 2.5 percentage points less likely
to be in grade 12 at age 17 than another sibling born under 40 weeks gestation.
Table 9 shows the estimated effects of infant health on the probability of
receiving any social assistance between the age of 18 and the last year for which we have
data (2004). Given that individuals in our birth cohorts were born between 1979 and
1985, only 8.9 percent of our sample ever received social assistance over this period. The
least-squares results without family fixed effects indicate a substantial relationship
between poor infant health and receiving social assistance. Apgar score, in particular,
strongly predicts take-up. A young adult given an Apgar of 6 at birth has a 5 percentage
point higher probability of receiving welfare than a contemporary given an Apgar of 10.
Individuals with Apgar scores between 7 and 9 also have a higher likelihood of receiving
social assistance. These effects are closer to zero once the family fixed effects are added
in column 3. The point estimates are somewhat noisy, but they suggest that the causal
effects of Apgar score on welfare use, independent of family circumstances that correlate
with this score, are lower than predicted by the strong effects estimated using the crossfamily variation. Interestingly, the estimates from the smaller sample of twins do suggest
a big effect. From column 5, a twin given an Apgar of 7 stands an 11.0 percentage point
higher probability of receiving social assistance than does her other twin with an Apgar
of 10.
The effects of birth weight on social assistance take-up also dissipate once family
fixed effects are included in the model. The least-squares results in column 2 indicate a
21
strong escalating relationship between being born low birth weight and receiving social
assistance. The point estimates fall substantially in column 3 from including the family
fixed effects. While, once again, the estimates for this outcome variable are imprecise,
the standard errors are small enough to rule out that the estimates are equal to those
without the fixed effects, while not ruling out the effects are zero. The twins results, like
the case with Apgar scores, are more suggestive of a long-lasting effect from infant
health.
The gestation results without fixed effects indicate a significant relationship
among youths born after less than 38 weeks gestation. These coefficients, however, all
drop close to zero (and are imprecisely measured) after adding in family fixed effects.
In an attempt to increase the variance of social assistance use in our sample, we
also look at months on social assistance between age 18 and (up to) age 25. Table 10
presents these results. The mean number of months on social assistance in our sample of
siblings is 2.1. The results are also somewhat noisy, but generally suggest a continued
link between our infant health measures and social assistance use. Birth weight appears
to affect not only take-up but also duration of social assistance. The coefficients for the
effects of low birth weight on months on social assistance using family fixed effects are
about one-half to two-thirds the size of those without the fixed effects. The estimates for
the average association between being born 1,500 to 3,000 grams and social assistance
use are significant, and we can reject the hypothesis that all the estimated effects are zero
or less. The coefficients from the twins sample with family fixed effects included, all
suggest a long-lasting effect of poor Apgar score or birth weight on months on social
assistance. The Apgar score results are not significant, but the implied effects of birth
22
weight are large, as they are when looking at only social assistance take-up (in the
previous table).
In Table 11, we reexamine our main sibling results for a sub-sample of siblings
less than 2 years and 3 years apart. A threat to validity in the siblings analysis comes
from changes to family or environmental contexts in between births that could account
for differences in socioeconomic outcomes. By looking at a sub-set of siblings closer in
age, fewer changes in family circumstances that may affect these outcomes are likely to
occur. The coefficients on the estimated effects of Apgar scores, birth weight, and
gestation on infant mortality largely remain intact after looking only at siblings less than
2 years apart.
The estimates with family fixed effects remain quite similar to the
estimates without them and indicate a strong relationship and suggest a significant causal
relationship between these measures of health at birth and one year mortality. It is worth
pointing out again that this contrasts with the twins results, where the estimated effects
fall by as much as two-thirds when accounting for family factors common between twins.
Table 12 shows the analysis for grade attainment outcomes between the sample
with all siblings and the one with siblings less than 2 years apart. The sub-sample is onefourth the size of the full sample. Yet, for birth weight, the results are remarkably stable.
Low birth weight siblings are approximately 10 percentage points less likely to attain
Grade 12 by age 17.
Even those born between 2500 and 3500 grams are 2 to 5
percentage points less likely to attain Grade 12 compared to those born weighing more
than 3500 grams. The estimated effects from being born premature or with a low Apgar
are measured less precisely with the sub-sample of siblings close apart, but the results
generally point to the same conclusions about impact of these measures on grade
23
attainment, with and without including family fixed effects. For more precision, we also
include the results for the sample of siblings less than 3 years apart in columns 5 and 6.
Low Apgar scores and gestation significantly impede high school grade progression.
Table 13 repeats the analysis, but for social assistance take-up outcomes. The full
sample results suggest significant effects of low birth weight on months receiving social
assistance after age 19, as does the sample of siblings less than 3 years apart. The
standard errors around the point estimates for the smaller sample of siblings less than 2
years apart prevent definitive conclusions. The Apgar results include a non-intuitive
results that those born with an Apgar score of 9 instead of 10 are slightly less likely to
end up on social assistance, while those born with an Apgar score of 7 or 8 are more
likely. None of the gestation results are significant.
We end our analysis by considering whether our estimated effects of infant health
on subsequent outcomes differ by family background. Columns 1 and 2 of Table 14
repeat the previous sibling sample results for the estimated effects of infant health on
one-year mortality, with and without family fixed effects. In the next two columns, we
show the results for the same model, but from using the sample of only births from
parents in residential areas where family incomes are among the first quintile. The
results among the bottom fifth group of families still reveal quite similar point estimates
compared to those from the full sample. For example, among children with parents from
the lowest residential income quintile, an Apgar score of 7 or 8 for one sibling is
associated with a 2.3 percentage point higher likelihood of death within one year than
another sibling given a 5 minute Apgar score of 10. Columns 5 and 6 include results for
the sample of births with parents from the bottom two residential income quintiles. This
24
increases the sample, while still focusing on families from poor social-economic
backgrounds. In general, we find no substantial differences in the estimated effects of
infant health on mortality comparing these more disadvantaged groups with the entire
population of births. This evidence is consistent with previous research (Currie and
Hyson, 1999) using an older cohort of births.
Table 15 shows the results by residential income quintile for whether an
individual born in poor health is more likely to have been held back or dropped out by
age 17. Here the imprecision of the estimates when using the smaller sample size of the
lower quintiles makes it more difficult to make comparisons with the full sample. The
point estimates for the lower quintile sample in columns 3 and 4 are higher than those for
the full sample in almost all cases, but not by much. In general, we conclude that there is
no strong evidence that the effects of infant health on high school attainment are any
worse among families from lower income backgrounds. The estimates with the two
lower quintiles also are not significantly different from the full sample results.
Finally, Table 16 displays the quintile results for months on social assistance.
The estimates are not suggestive of any differences between the lower quintile groups
and the population. However, the estimates are somewhat imprecise. Overall, our results
by quintile suggest that the short, medium, and longer-term effects of infant health are not
confined to a single quintile, but rather are uniform across the population.
25
5. CONCLUSIONS
We use a cohort of births from a single Canadian province to examine the short,
medium and long-term effects of poor infant health. Our results both confirm and extend
recent work on the effects of infant health on survival and future measures of health,
human capital, and labor force attachment.
Using three measures of infant health: birth weight, Apgar scores, and gestational
length, we find that poor infant health predicts both mortality within 1-year, and mortality
up to age 17. These results hold both across families and between siblings within
families. Consistent with results in Almond et al. (2005), differences in infant health
within families but between twins pairs lead to much smaller differences in both 1 year
and 17 year mortality rates. This drop in the estimated effects occurs for twins but not
for siblings.
We also find that infant health is a strong predictor of educational and labor force
outcomes. In particular, infant health is found to predict both high school completion and
social assistance (welfare) take-up and length. We find evidence of longer-term
consequences of infant health both across families, within siblings, and within twin pairs,
although different measures of infant health predict outcomes differently. Interestingly,
we find less evidence of the longer-term effects of infant health on either cognitive ability
as measured by language arts test scores or longer-term physician visits.
Our evidence, along with a growing body of literature in this area, confirms the
importance of early childhood health as a predictor of future outcomes. Examining
differences across families, between siblings, and between twin pairs can help inform
26
both the medical literature and public policy with regards to the most effective way to
improve childhood health and hence future outcomes.
27
REFERENCES:
Almond, D, Chay, K., and David Lee, “The Costs of Low Birth Weight”, Quarterly
Journal of Economics, 120(3), 2005.
Behrman, J., and Mark Rosenzweig, “The Returns to Increasing Body Weight,” Penn
Institute for Economic Research Working Paper 01-052, 2001.
Behrman, J., and Mark Rosenzweig, “The Returns to Birth Weight,” Review of
Economics and Statistics, 86, 2004.
Black, S. Devereux, P., Kjell Salvanes, “From the Cradle to the Labor Market: The Effect
of Birth Weight on Adult Outcomes,” NBER Working Paper 11796, 2005.
Black, S., Devereux, P., and Kjell Salvanes, “The More the Merrier? The Effect of
Family Size and Birth Order on Children’s Education,” The Quarterly Journal of
Economics, May, 2005.
Conley, D., Strullly, K., and Neil Bennett, “A Pound of Flesh or Just Proxy? Using Twin
Differences To Estimate The Effect of Birth Weight on Life Chances,” NBER Working
Paper 9901, 2003.
Conley, D., Strullly, K., and Neil Bennett, The Starting Gate: Birth Weight and Life
Chances, Berkeley, CA: University of California Press, 2003.
Currie, Janet, and Gruber, Jonathan . “Health Insurance Eligibility, Utilization of Medical
Care, and Child Health,” Quarterly Journal of Economics, vol. 111, no. 2, May 1996, pp.
431-66
Currie, Janet, and Moretti, Enrico. “Biology as Destiny? Short and Long-Run
Determinants of Intergenerational Transmission of Birth Weight,” NBER Working Paper
#11567.
Currie, Janet, and Rosemary Hyson, “"Is the Impact of Health Shocks Cushioned by
Socioeconomic Status?: The Case of Birth Weight," American Economic Review, May
1999, 89 #2, 245-250.
Goldenberg, R., and Dwight Rouse, “Prevention of Premature Birth,” New England
Journal of Medicine, 339, 1998.
Hanratty, M. “Canadian National Health Insurance and Infant Health,” American
Economic Review, 86(1), March, 1996.
Kramer, M. “Intrauterine Growth and Gestational Duration Determinants,” Pediatrics,
58, 1987.
29
Mosteller F, Tukey JW. Data Analysis and Regression. A Second Course in Statistics.
Reading, MA: Addison-Wesley, 1977.
Mustard, C., Derksen, S., Berthelot, J.M., Wolfson, M. "Assessing ecologic proxies for
household income: a comparison of household and neighbourhood level income measures
in the study of population health status," Health and Place 5:2 (1999) 157-171.
National Institutes of Health, “NIH Guide” Low Birth Weight in Minority Populations,”
PA-99-045, 1999.
Roos, LL., Walld, R., Uhanova, J. and R. Bond, "Physician visits, hospitalizations, and
socioeconomic status: ambulatory care sensitive conditions in a Canadian setting," Health
Services Research 40:4 (August 2005), 1167-1185.
Roos LL., and JP Nicol, "A research registry: uses, development, and accuracy," J Clin
Epidemiology 52:1 (1999), 39-47.
Watson DE, Katz A, Reid RJ, Bogdanovic B, Roos N, Heppner P. Family physician
workloads and access to care in Winnipeg: 1991 to 2001. Canadian Medical Association
Journal. 2004;171(4):339–42.
Willms, J.D, “Social class segregation and its relationship to pupils' examination results
in Scotland,” American Sociological Review, 1986;51(2):224-241.
30
Table 1
Descriptive Statistics of Infant Health Measures
5 Minute APGAR Score (0-10)
Sample
Mean
s.d.
Within Family
s.d.
1st
5th
10th
Percentile
25th
50th
75th
N
All Births 1979-85
9.094
0.928
NA
6
8
8
9
9
10
108800
Siblings Only
9.131
0.919
0.646
6
8
8
9
9
10
54123
Twins Only
8.594
1.397
0.625
1
6
7
9
9
9
1742
Birth Weight (in Grams)
Mean
s.d.
Within Family
s.d.
1st
5th
10th
Percentile
25th
50th
75th
N
All Births 1979-85
3424.5
570.3
NA
1680
2500
2760
3100
3450
3780
109125
Siblings Only
3458.3
556.6
314.1
1780
2580
2820
3140
3480
3800
54986
Twins Only
2517.4
610.6
201.9
790
1360
1725
2181.5
2580
2930
1752
Sample
Gestation (in Weeks)
Sample
Mean
s.d.
Within Family
s.d.
1st
5th
10th
Percentile
25th
50th
75th
N
All Births 1979-85
39.4
2.0
NA
32
36
37
39
40
40
90135
Siblings Only
39.4
1.9
1.1
33
37
38
39
40
40
45583
Twins Only
36.5
3.3
0.0
25
30
32
35
37
39
1492
Table 2
Descriptive Statistics of Outcome Measures (Sibling Sample)
1979 - 1985 Manitoba Births
Mean
s.d.
Age of
Individual
N
Infant Mortality
0.011
0.105
To 365 days
54310
Death between ages 1 and 17
0.006
0.080
17
53700
Moved from Manitoba
0.208
0.406
17
53750
Total Physician Visits
14.358
12.610
Age 12-17
40203
Language Score (standardized scaled logit)
-0.016
1.013
Grade 12
40203
Reached Grade 12 by Age 17
0.694
0.461
17
40203
Ever on Social Assistance
0.089
0.285
Age 18 to
Mar-04
40203
Months on Social Assistance
2.060
9.343
Age 18 to
Mar-04
40203
Table 3
Estimated Effects of Infant Health at Birth on Infant Mortality (Death within One Year of Birth)
With and Without Family Fixed Effects
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
With Family F.E.
Twins Sample
No Family F.E.
Twins Sample
With Family F.E.
APGAR Score (Omitted Category APGAR>6)
APGAR<=6
0.2588 ***
(0.0021)
0.3092 ***
(0.0034)
0.3197 ***
(0.0047)
0.3123 ***
(0.0171)
0.0957 ***
(0.0237)
APGAR=7-8
0.0163 ***
(0.0010)
0.0189 ***
(0.0015)
0.0198 ***
(0.0022)
0.0186 *
(0.0111)
-0.0045
(0.0155)
APGAR=9
0.0017 ***
(0.0006)
0.0025 ***
(0.0009)
0.0020
(0.0014)
0.0024
(0.0097)
-0.0014
(0.0147)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
5036.06 ***
{108893}
0.12
2887.21 ***
{54091}
0.13
1588.14 ***
{54091}
0.50
137.49
{1740}
0.34
***
8.93
{1740}
0.84
***
Birth Weight (Omitted Category BW = 2500-3000 grams)
BW <1000
0.8120 ***
(0.0046)
0.8572 ***
(0.0071)
0.8723 ***
(0.0099)
0.7249 ***
(0.0292)
0.2532 ***
(0.0701)
BW 1000-1500
0.2657 ***
(0.0039)
0.3622 ***
(0.0066)
0.3912 ***
(0.0093)
0.2099 ***
(0.0220)
0.0848 **
(0.0384)
BW 1500-2500
0.0320 ***
(0.0013)
0.0479 ***
(0.0022)
0.0630 ***
(0.0034)
0.0051
(0.0174)
0.0048
(0.0246)
BW 2500-3000
0.0058 ***
(0.0008)
0.0065 ***
(0.0012)
0.0130 ***
(0.0021)
0.0063
(0.0174)
-0.0043
(0.0231)
BW 3000-3500
0.0015 ***
(0.0006)
0.0018 **
(0.0009)
0.0042 ***
(0.0014)
-0.0006
(0.0182)
-0.0017
(0.0218)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
7099.73 ***
{109114}
0.05
3586.36 ***
{54986}
0.06
1899.91 ***
{54986}
0.49
213.85
{2020}
0.24
***
3.25
{2020}
0.86
***
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
0.0837 ***
(0.0014)
0.1060 ***
(0.0023)
0.1187 ***
(0.0039)
0.0804 ***
(0.0104)
NA
Gestation 37 weeks
0.0056 ***
(0.0016)
0.0084 ***
(0.0025)
0.0142 ***
(0.0041)
0.0060
(0.0141)
NA
Gestation 38 weeks
0.0035 ***
(0.0010)
0.0052 ***
(0.0015)
0.0073 ***
(0.0026)
0.0044
(0.0123)
NA
Gestation 39 weeks
0.0010
(0.0008)
0.0005
(0.0013)
-0.0003
(0.0021)
0.0089
(0.0132)
NA
Gestation >=42 weeks
0.0005
(0.0013)
-0.0003
(0.0019)
0.0005
(0.0031)
0.0001
(0.0237)
NA
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
722.48 ***
{90460}
0.09
435.15 ***
{46183}
0.10
198.70 ***
{46183}
0.55
7.55
{1976}
0.29
***
NA
NA
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies
for the birth order of the child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and
1 percent levels respectively.
Table 4
Estimated Effects of Infant Health at Birth on Mortality between Ages 1 to 17
With and Without Family Fixed Effects
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
With Family F.E.
Twins Sample
No Family F.E.
Twins Sample
With Family F.E.
APGAR Score (Omitted Category APGAR=10)
APGAR<=6
0.0111 ***
(0.0020)
0.0147 ***
(0.0033)
0.0131 ***
(0.0045)
0.0100
(0.0076)
-0.0159
(0.0136)
APGAR=7-8
0.0026 ***
(0.0008)
0.0038 ***
(0.0013)
0.0034 *
(0.0018)
-0.0071 *
(0.0043)
-0.0150 *
(0.0087)
APGAR=9
0.0001
(0.0005)
-0.0002
(0.0008)
0.0004
(0.0011)
-0.0040
(0.0038)
-0.0092
(0.0082)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
14.1 ***
{107772}
0.01
10.4 ***
{53527}
0.01
3.88 ***
{53527}
0.48
2.31 *
{1693}
0.07
1.13
{1693}
0.55
Birth Weight (Omitted Category BW = 3500+ grams)
BW <1000
0.0444 ***
(0.0094)
BW 1000-1500
0.0124 ***
(0.0040)
BW 1500-2500
0.0025 **
(0.0011)
BW 2500-3000
-0.0058
(0.0169)
-0.0384 *
(0.0226)
0.0004
(0.0216)
0.0170
(0.0488)
0.0291 ***
(0.0101)
0.0161
(0.0100)
0.0169
(0.0211)
0.0043 **
(0.0020)
0.0068 **
(0.0030)
0.0002
(0.0075)
0.0013
(0.0132)
0.0003
(0.0007)
0.0006
(0.0011)
0.0019
(0.0018)
0.0010
(0.0075)
0.0028
(0.0124)
BW 3000-3500
0.0006
(0.0005)
0.0009
(0.0008)
0.0034 ***
(0.0012)
0.0043
(0.0078)
-0.0013
(0.0117)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
7.36 ***
{108073}
0.01
2.44 **
{53700}
0.01
4.24 ***
{53700}
0.48
1.15
{1700}
0.06
0.26
{1700}
0.55
0.0193 ***
(0.0073)
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
0.0028 **
(0.0011)
0.0024
(0.0019)
0.0039
(0.0030)
0.0023
(0.0049)
NA
Gestation 37 weeks
0.0010
(0.0012)
0.0009
0.001918
0.0012
(0.0030)
-0.0027
(0.0060)
NA
Gestation 38 weeks
0.0004
(0.0008)
0.0014
(0.0012)
0.0027
(0.0019)
-0.0029
(0.0056)
NA
Gestation 39 weeks
0.0003
(0.0006)
0.0014
(0.0012)
0.0008
(0.0015)
0.0020
(0.0059)
NA
Gestation >=42 weeks
0.0008
(0.0010)
1.38
{89276}
0.01
-0.0004
(0.0010)
0.71
{45078}
0.01
-0.0002
(0.0023)
0.63
{45078}
0.54
-0.0068
(0.0137)
0.47
{1447}
0.08
NA
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies
for the birth order of the child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and
1 percent levels respectively.
Table 5
Estimated Effects of Infant Health at Birth on Mobility out of Manitoba Before Age 18
With and Without Family Fixed Effects
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
With Family F.E.
Twins Sample
No Family F.E.
Twins Sample
With Family F.E.
APGAR Score (Omitted Category APGAR=10)
APGAR<=6
0.0241 **
(0.0115)
APGAR=7-8
0.0120 ***
(0.0046)
APGAR=9
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
0.0169
(0.0165)
-0.0195 *
(0.0110)
0.0465
(0.0548)
0.0169
(0.0178)
0.0072
(0.0062)
0.0023
(0.0043)
0.0470
(0.0308)
0.0111
(0.0113)
0.0042
(0.0030)
0.0067 *
(0.0038)
-0.0002
(0.0028)
0.0174
(0.0270)
0.0148
(0.0107)
3.38 **
{107772}
0.07
1.29
{53527}
0.05
1.24
{53527}
0.88
0.94
{1693}
0.06
0.70
{1693}
0.99
Birth Weight (Omitted Category BW = 3500+ grams)
BW <1000
-0.0016
(0.0545)
-0.0161
(0.0836)
-0.0031
(0.0552)
0.2391
(0.1548)
-0.0348
(0.0634)
BW 1000-1500
-0.0175
(0.0230)
-0.00908
(0.0361)
0.0041
(0.0245)
0.0564
(0.0716)
-0.0253
(0.0274)
BW 1500-2500
0.0126 *
(0.0066)
0.0271 ***
(0.0098)
0.0027
(0.0073)
0.1148 **
(0.0536)
-0.0249
(0.0171)
BW 2500-3000
0.0099 **
(0.0039)
0.0079
(0.0053)
-0.0045
(0.0043)
0.0988 *
(0.0537)
-0.0314 *
(0.0161)
BW 3000-3500
0.0058 **
(0.0029)
0.0061
(0.0038)
0.0027
(0.0029)
0.1307 **
(0.0560)
-0.0188
(0.0152)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
2.11 *
{108073}
0.07
1.97 *
{53700}
0.05
0.80
{53700}
0.88
1.58
{1700}
0.06
0.96
{1700}
0.99
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
0.0076
(0.0063)
Gestation 37 weeks
0.0122 *
(0.0070)
Gestation 38 weeks
-0.0090 **
(0.0045)
Gestation 39 weeks
Gestation >=42 weeks
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
0.0157 *
(0.0092)
0.0102
(0.0069)
0.0511
(0.0318)
NA
0.0172 *
(0.0095)
-0.0059
(0.0070)
0.0092
(0.0392)
NA
-0.0022
(0.0059)
0.0055
(0.0045)
0.0644 *
(0.0367)
NA
-0.0008
(0.0037)
-0.0010
(0.0048)
-0.0003
(0.0036)
0.0589
(0.0386)
NA
0.0044
(0.0056)
2.12 *
{89276}
0.06
0.0042
(0.0073)
1.36
{45078}
0.04
-0.0033
(0.0052)
1.12
{45078}
0.90
-0.1367
(0.0894)
1.93 *
{1447}
0.06
NA
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies for
the birth order of the child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and 1
percent levels respectively.
Table 6
Estimated Effects of Infant Health at Birth on Total Physician Visits Between Ages 12 and 17
With and Without Family Fixed Effects
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
APGAR Score (Omitted Category APGAR=10)
APGAR<=6
1.9767 ***
(0.4213)
1.5734 **
(0.6186)
1.5356 **
(0.6768)
1.9612
(1.7188)
0.7656
(2.2051)
APGAR=7-8
0.5790 ***
(0.1629)
0.2692
(0.2243)
-0.5566 **
(0.2561)
-1.0923
(0.9297)
-1.2239
(1.3159)
APGAR=9
0.3954 ***
(0.1043)
0.3173 **
(0.1376)
-0.1826
(0.1628)
0.0717
(0.8024)
-0.4005
(1.2520)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
11.76 ***
{79143}
0.04
3.47
**
{40078}
0.05
3.81
***
{40078}
0.68
1.58
{1348}
0.09
0.74
{1348}
0.78
Birth Weight (Omitted Category BW = 3500+ grams)
BW <1000
-0.9530
(1.9892)
-0.8204
(3.1370)
2.2148
(3.2498)
2.6851
(4.8230)
4.9343
(10.6648)
BW 1000-1500
1.0137
(0.8185)
-0.5904
(1.3344)
-1.6869
(1.4809)
9.3137 ***
(2.1409)
7.3252 **
(2.9635)
BW 1500-2500
0.4262 *
(0.2359)
-0.1540
(0.3662)
0.5461
(0.4425)
3.3341 **
(1.5216)
0.1926
(1.8861)
BW 2500-3000
0.2481 *
(0.1402)
0.3470 *
(0.1941)
0.2464
(0.2537)
3.5727 **
(1.5280)
0.1849
(1.7688)
BW 3000-3500
0.2125 **
(0.1027)
0.3026 **
(0.1375)
0.3215 *
(0.1704)
2.9624 *
(1.6084)
1.9898
(1.6555)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
1.78
{79363}
0.04
1.47
{40203}
0.05
1.25
{40203}
0.68
4.03
{1354}
0.10
***
2.77
{1354}
0.78
**
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
0.2814
(0.2271)
-0.2208
(0.3395)
-0.5687
(0.4291)
0.8215
(0.9276)
NA
Gestation 37 weeks
0.3913
(0.2491)
0.1131
(0.3487)
-0.1167
(0.4345)
3.1315 ***
(1.1326)
NA
Gestation 38 weeks
0.4832 ***
(0.1582)
0.2030
(0.2119)
0.0325
(0.2693)
-0.1308
(1.0700)
NA
Gestation 39 weeks
0.0867
(0.1305)
-0.1401
(0.1731)
-0.1719
(0.2151)
-0.0701
(1.1335)
NA
Gestation >=42 weeks
0.0874
(0.1973)
-0.0581
(0.2636)
-0.0985
(0.3201)
-1.3445
(2.3796)
NA
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
2.29
**
{66504}
0.04
0.53
{33921}
0.05
0.48
{33921}
0.72
2.59
{1166}
0.10
**
NA
NA
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies
for the birth order of the child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and
1 percent levels respectively.
Table 7
Estimated Effects of Infant Health at Birth on Language Arts Score
With and Without Family Fixed Effects
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
APGAR Score (Omitted Category APGAR>6)
APGAR<=6
-0.1541 ***
(0.0297)
-0.1730 ***
(0.0450)
-0.0913 *
(0.0481)
0.1254
(0.1457)
0.0756
(0.1568)
APGAR=7-8
-0.0503 ***
(0.0116)
-0.0436 ***
(0.0164)
-0.0245
(0.0183)
0.0326
(0.0786)
-0.1021
(0.0936)
APGAR=9
0.0069
(0.0074)
0.0184 *
(0.0101)
-0.0012
(0.0116)
0.0444
(0.0676)
-0.0952
(0.0890)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
18.52 ***
{79194}
0.17
10.97 ***
{40514}
0.20
1.82
{40514}
0.75
0.30
{1364}
0.20
0.99
{1364}
0.87
Birth Weight (Omitted Category BW = 2500-3000 grams)
BW <1000
-0.31424 **
0.141101
-0.25756
0.230887
-0.04906
0.231668
-0.54153
0.408117
0.01819
0.761597
BW 1000-1500
-0.26002 ***
0.058061
-0.2302 **
0.098216
-0.08117
0.105565
-0.46083 **
0.181157
-0.18498
0.211632
BW 1500-2500
-0.1519 ***
0.016731
-0.21882 ***
0.026951
-0.04925
0.031547
-0.18261
0.128753
-0.09871
0.134692
BW 2500-3000
-0.1172 ***
(0.0099)
-0.1303 ***
(0.0143)
-0.0477 **
(0.0181)
-0.2063
(0.1293)
-0.2139 *
(0.1263)
BW 3000-3500
-0.0339 ***
(0.0073)
-0.0459 ***
(0.0101)
-0.0154
(0.0121)
-0.0969
(0.1361)
-0.0672
(0.1182)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
42.74 ***
{79363}
0.17
27.73 ***
{40203}
0.20
1.57
{40203}
0.74
1.95
{1354}
0.21
*
1.54
{1354}
0.86
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
-0.0797 ***
(0.0161)
-0.1232 ***
(0.0251)
0.0255
(0.0303)
0.0061
(0.0800)
NA
Gestation 37 weeks
-0.0678 ***
(0.0177)
-0.0999 ***
(0.0258)
-0.0076
(0.0307)
0.1317
(0.0976)
NA
Gestation 38 weeks
-0.0114
(0.0112)
-0.0166
(0.0157)
0.0263
(0.0190)
-0.0708
(0.0922)
NA
Gestation 39 weeks
0.0083
(0.0093)
0.0003
(0.0128)
0.0037
(0.0152)
0.0543
(0.0977)
NA
Gestation >=42 weeks
-0.0204
(0.0140)
-0.0231
(0.0195)
-0.0089
(0.0226)
-0.1789
(0.2051)
NA
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
8.44
***
{66504}
0.17
7.70
***
{33921}
0.20
0.60
{33921}
0.79
1.28
{1166}
0.21
NA
NA
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies for
the birth order of the child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and 1
percent levels respectively.
Table 8
Estimated Effects of Infant Health at Birth on Reaching Grade 12 by Age 17
With and Without Family Fixed Effects
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
APGAR Score (Omitted Category APGAR>6)
APGAR<=6
-0.0697 ***
(0.0137)
-0.0739 ***
(0.0206)
-0.0410 *
(0.0239)
0.0710
(0.0701)
0.0317
(0.0706)
APGAR=7-8
-0.0240 ***
(0.0053)
-0.0302 ***
(0.0075)
-0.0203 **
(0.0091)
0.0179
(0.0378)
-0.0210
(0.0421)
APGAR=9
0.0058 *
(0.0034)
0.0111 **
(0.0046)
0.0051
(0.0058)
0.0428
(0.0325)
-0.0156
(0.0401)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
21.3
***
{79194}
0.15
16.3
***
{40514}
0.18
4.12
***
{40514}
0.70
0.81
{1364}
0.18
0.32
{1364}
0.88
Birth Weight (Omitted Category BW = 2500-3000 grams)
BW <1000
-0.1842 ***
(0.0649)
-0.1308
(0.1060)
0.0546
(0.1153)
-0.3346 *
(0.1968)
-0.0837
(0.3434)
BW 1000-1500
-0.1353 ***
(0.0267)
-0.1271 ***
(0.0451)
-0.1410 ***
(0.0525)
-0.2708 ***
(0.0874)
-0.2315 **
(0.0954)
BW 1500-2500
-0.0793 ***
(0.0077)
-0.1168 ***
(0.0124)
-0.0815 ***
(0.0157)
-0.1284 **
(0.0621)
-0.0685
(0.0607)
BW 2500-3000
-0.0398 ***
(0.0046)
-0.0496 ***
(0.0066)
-0.0430 ***
(0.0090)
-0.1094 *
(0.0624)
-0.0921
(0.0570)
BW 3000-3500
-0.0125 ***
(0.0034)
-0.0189 ***
(0.0046)
-0.0243 **
(0.0060)
-0.0639
(0.0656)
-0.0488
(0.0533)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
37.47 ***
{79363}
0.16
27.74 ***
{40203}
0.18
8.98
***
{40203}
0.69
2.91
{1354}
0.19
**
1.72
{1354}
0.88
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
-0.0602 ***
(0.0074)
-0.0843 ***
(0.0115)
-0.0403 ***
(0.0150)
-0.0498
(0.0387)
NA
Gestation 37 weeks
-0.0246 ***
(0.0081)
-0.0426 ***
(0.0118)
-0.0345 **
(0.0152)
0.0473
(0.0472)
NA
Gestation 38 weeks
-0.0064
(0.0051)
-0.0139 *
(0.0072)
-0.0252 ***
(0.0094)
0.0419
(0.0446)
NA
Gestation 39 weeks
0.0076 *
(0.0042)
0.0062
(0.0059)
-0.0014
(0.0075)
0.0580
(0.0473)
NA
Gestation >=42 weeks
-0.0067
(0.0064)
-0.0099
(0.0089)
-0.0116
(0.0112)
-0.1740 *
(0.0992)
NA
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
17.14 ***
{66504}
0.16
14.37 ***
{33921}
0.19
3.19
***
{33921}
0.75
3.23
{1166}
0.19
***
NA
NA
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies for
the birth order of the child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and 1
percent levels respectively.
Table 9
Estimated Effects of Infant Health at Birth on Social Assistance Take-up
With and Without Family Fixed Effects
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
APGAR Score (Omitted Category APGAR>6)
APGAR<=6
0.0531 ***
(0.0090)
0.0513 ***
(0.0136)
0.0166
(0.0162)
-0.0187
(0.0422)
0.0518
(0.0600)
APGAR=7-8
0.0213 ***
(0.0035)
0.0223 ***
(0.0049)
0.0030
(0.0061)
0.0065
(0.0228)
0.1102 ***
(0.0358)
APGAR=9
0.0119 ***
(0.0022)
0.0105 ***
(0.0030)
-0.0058
(0.0039)
-0.0077
(0.0197)
0.0896 ***
(0.0340)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
23.75 ***
{79143}
0.08
11.34 ***
{40078}
0.10
1.81
{40078}
0.64
0.29
{1348}
0.09
3.46 *
{1348}
0.73
Birth Weight (Omitted Category BW = 3500+ grams)
BW <1000
0.0076
(0.0424)
-0.0602
(0.0691)
-0.1444 *
(0.0775)
0.1720
(0.1192)
0.0529
(0.2951)
BW 1000-1500
0.0696 ***
(0.0174)
0.0472
(0.0294)
0.0263
(0.0353)
0.1385 ***
(0.0529)
0.0932
(0.0820)
BW 1500-2500
0.0366 ***
(0.0050)
0.0387 ***
(0.0081)
0.0119
(0.0106)
0.0702 *
(0.0376)
0.0583
(0.0522)
BW 2500-3000
0.0216 ***
(0.0030)
0.0237 ***
(0.0043)
0.0031
(0.0061)
0.0462
(0.0378)
0.0800
(0.0489)
BW 3000-3500
0.0058 ***
(0.0022)
0.0078 ***
(0.0030)
-0.0003
(0.0041)
0.0697 *
(0.0397)
0.0990 **
(0.0458)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
21.15 ***
{79363}
0.08
10.20 ***
{40203}
0.10
1.12
{40203}
0.64
2.08 *
{1354}
0.10
1.22
{1354}
0.73
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
0.0263 ***
(0.0048)
0.02938 ***
0.007427
0.0017
(0.0101)
-0.0044
(0.0225)
NA
Gestation 37 weeks
0.0125 **
(0.0052)
0.017372 **
0.00763
-0.0087
(0.0102)
-0.0274
(0.0275)
NA
Gestation 38 weeks
0.0058 *
(0.0033)
0.007717 *
0.004637
0.0027
(0.0063)
-0.0212
(0.0260)
NA
Gestation 39 weeks
-0.0019
(0.0027)
-0.00056
0.003787
-0.0017
(0.0050)
-0.0533 *
(0.0275)
NA
Gestation >=42 weeks
0.0043
(0.0042)
0.002119
0.005768
-0.0098
(0.0075)
-0.0413
(0.0577)
NA
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
7.86 ***
{66504}
0.08
4.43 ***
{33921}
0.09
0.60
{33921}
0.70
1.16
{1166}
0.10
NA
NA
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies for
the birth order of the child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and 1
percent levels respectively.
Table 10
Estimated Effects of Infant Health at Birth on Months on Social Assistance
With and Without Family Fixed Effects
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
APGAR Score (Omitted Category APGAR>6)
APGAR<=6
2.0171 ***
(0.2941)
1.7872 ***
(0.4518)
0.8932
(0.5716)
0.8242
(1.3382)
1.7657
(2.0349)
APGAR=7-8
0.7140 ***
(0.1137)
0.6922 ***
(0.1638)
0.2739
(0.2163)
0.8457
(0.7239)
1.9836
(1.2143)
APGAR=9
0.3255 ***
(0.0728)
0.2390 **
(0.1005)
-0.2753 **
(0.1375)
0.4264
(0.6248)
1.5359
(1.1553)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
26.52 ***
{79143}
0.06
10.32 ***
{40078}
0.07
4.36 ***
{40078}
0.58
0.49
{1348}
0.10
0.89
{1348}
0.69
Birth Weight (Omitted Category BW = 3500+ grams)
BW <1000
-0.3353
(1.3897)
-2.3661
(2.2936)
-1.4601
(2.7460)
2.0097
(3.7608)
-0.2773
(9.8061)
BW 1000-1500
2.4388 ***
(0.5719)
0.6699
(0.9757)
0.3227
(1.2513)
2.7609 *
(1.6694)
3.3494
(2.7249)
BW 1500-2500
1.2755 ***
(0.1648)
1.5665 ***
(0.2677)
1.0286 ***
(0.3739)
1.1299
(1.1865)
3.7207 **
(1.7343)
BW 2500-3000
0.6888 ***
(0.0980)
0.8635 ***
(0.1419)
0.4263 **
(0.2144)
0.8380
(1.1915)
4.6565 ***
(1.6264)
BW 3000-3500
0.1962 ***
(0.0717)
0.2678 ***
(0.1006)
0.1816
(0.1440)
2.4619 **
(1.2542)
5.5686 ***
(1.5222)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
22.26 ***
{79363}
0.06
13.30 ***
{40203}
0.07
1.88 *
{40203}
0.58
1.83
{1354}
0.10
3.08 ***
{1354}
0.70
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
0.7807 ***
(0.1400)
0.803194 ***
0.222536
0.2854
(0.3192)
0.2894
(0.5645)
NA
Gestation 37 weeks
0.4095 ***
(0.1536)
0.459056 **
0.228595
0.1549
(0.3233)
-0.5299
(0.6892)
NA
Gestation 38 weeks
0.1784 *
(0.0975)
0.183002
0.138931
0.0710
(0.2004)
-0.0699
(0.6511)
NA
Gestation 39 weeks
-0.1356 *
(0.0805)
-0.13778
0.113471
-0.0582
(0.1601)
-0.6536
(0.6898)
NA
Gestation >=42 weeks
0.1612
(0.1217)
0.18277
0.172823
-0.0483
(0.2382)
-0.0180
(1.4480)
NA
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
9.54 ***
{66504}
0.06
4.42 ***
{33921}
0.07
0.28
{33921}
0.65
0.75
{1166}
0.13
NA
NA
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies for
the birth order of the child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and 1
percent levels respectively.
Table 11
Estimated Effects of Infant Health at Birth on Infant Mortality
Close Siblings as Compared to All Siblings, With and Without Family Fixed Effects
Whole Sibling Sample
Sibling Sample
Sibling Sample
No Family F.E.
Family F.E.
Born Within 2 Years
Sibling Sample
Sibling Sample
No Family F.E.
Family F.E.
Born Within 3 Years
Sibling Sample
Sibling Sample
No Family F.E.
Family F.E.
APGAR<=6
0.3092 ***
(0.0034)
0.3197 ***
(0.0047)
0.3732 ***
(0.0061)
0.3773 ***
(0.0089)
0.3105
(0.0042)
***
0.3157
(0.0061)
***
APGAR=7-8
0.0189 ***
(0.0015)
0.0198 ***
(0.0022)
0.0231 ***
(0.0030)
0.0240 ***
(0.0045)
0.0191
(0.0019)
***
0.0193
(0.0029)
***
APGAR=9
0.0025 ***
(0.0009)
0.0020
(0.0014)
0.0026
(0.0018)
0.0028
(0.0029)
0.0019
(0.0012)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
2887.21 ***
{54091}
0.13
1588.14 ***
{54091}
0.50
1292.22 ***
{17948}
0.19
625.82 ***
{17948}
0.59
1856.15
{34946}
0.15
***
915.34
{34946}
0.57
***
BW <1000
0.8572 ***
(0.0071)
0.8723 ***
(0.0099)
0.8947 ***
(0.0122)
0.9275 ***
(0.0176)
0.8555
(0.0087)
***
0.8756
(0.0127)
***
BW 1000-1500
0.3622 ***
(0.0066)
0.3912 ***
(0.0093)
0.4199 ***
(0.0111)
0.4338 ***
(0.0160)
0.3872
(0.0079)
***
0.4054
(0.0115)
***
BW 1500-2500
0.0479 ***
(0.0022)
0.0630 ***
(0.0034)
0.0598 ***
(0.0041)
0.0731 ***
(0.0066)
0.0513
(0.0028)
***
0.0640
(0.0044)
***
BW 2500-3000
0.0065 ***
(0.0012)
0.0130 ***
(0.0021)
0.0034
(0.0024)
0.0090 **
(0.0041)
0.0032
(0.0015)
**
0.0088
(0.0026)
***
BW 3000-3500
0.0018 **
(0.0009)
0.0042 ***
(0.0014)
0.0022
(0.0017)
0.0058 **
(0.0028)
0.0022
(0.0011)
**
0.0034
(0.0018)
*
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
3586.36 ***
{54310}
0.25
1899.91 ***
{54310}
0.57
1391.87 ***
{18036}
0.29
705.33 ***
{18036}
0.64
2449.40
{35090}
0.27
***
1205.97
{35090}
0.63
***
Gestation<=36 weeks
0.1060 ***
(0.0023)
0.1187 ***
(0.0039)
0.1214 ***
(0.0041)
0.1427 ***
(0.0073)
0.1073
(0.0028)
***
0.1209
(0.0049)
***
Gestation 37 weeks
0.0084 ***
(0.0025)
0.0142 ***
(0.0041)
0.0079 *
(0.0047)
0.0244 ***
(0.0078)
0.0054
(0.0031)
*
0.0140
(0.0051)
***
Gestation 38 weeks
0.0052 ***
(0.0015)
0.0073 ***
(0.0026)
0.0074 **
(0.0030)
0.0104 **
(0.0051)
0.0044
(0.0019)
**
0.0049
(0.0033)
Gestation 39 weeks
0.0005
(0.0013)
-0.0003
(0.0021)
-0.0024
(0.0025)
-0.0073 *
(0.0041)
0.0004
(0.0016)
-0.0037
(0.0026)
Gestation >=42 weeks
-0.0003
(0.0019)
0.0005
(0.0031)
-0.0015
(0.0038)
0.0007
(0.0061)
0.0005
(0.0024)
0.0008
(0.0039)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
435.15 ***
{45583}
0.05
198.70 ***
{45583}
0.51
183.06 ***
{15644}
0.07
84.88 ***
{15644}
0.58
304.47
{30218}
0.06
All regression models include additional fixed effects for mother's martial status, gender of child, and
family sibling size dummies for the birth order of the child within each family size. One, two, and three
asteriks indicate statistical significance at the 10, 5, and 1 percent levels respectively.
0.0023
(0.0018)
***
132.44
{30218}
0.58
***
Table 12
Estimated Effects of Infant Health at Birth on Reaching Grade 12 by Age 17
Close Siblings as Compared to All Siblings, With and Without Family Fixed Effects
Whole Sibling Sample
Sibling Sample
Sibling Sample
No Family F.E.
Family F.E.
Born Within 2 Years
Sibling Sample
Sibling Sample
No Family F.E.
Family F.E.
Born Within 3 Years
Sibling Sample
Sibling Sample
No Family F.E.
Family F.E.
APGAR<=6
-0.0739 ***
(0.0206)
-0.0410 *
(0.0239)
-0.0493
(0.0372)
-0.0137
(0.0437)
-0.0750
(0.0257)
***
-0.0509
(0.0306)
*
APGAR=7-8
-0.0302 ***
(0.0075)
-0.0203 **
(0.0091)
-0.0498 ***
(0.0135)
-0.0329 *
(0.0168)
-0.0368
(0.0093)
***
-0.0265
(0.0117)
**
APGAR=9
0.0111 **
(0.0046)
0.0051
(0.0058)
-0.0056
(0.0084)
-0.0166
(0.0108)
0.0093
(0.0057)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
16.30 ***
{40514}
0.18
4.12
***
{40514}
0.70
5.12
***
{12929}
0.19
1.45
{12929}
0.74
12.09
{26204}
0.18
BW <1000
-0.1308
(0.1060)
0.0546
(0.1153)
-0.0066
(0.2486)
0.5569 *
(0.2849)
-0.0238
(0.1313)
BW 1000-1500
-0.1271 ***
(0.0451)
-0.1410 ***
(0.0525)
-0.1154
(0.0717)
-0.0691
(0.0826)
-0.1069
(0.0546)
*
-0.0931
(0.0652)
BW 1500-2500
-0.1168 ***
(0.0124)
-0.0815 ***
(0.0157)
-0.1169 ***
(0.0221)
-0.0940 ***
(0.0289)
-0.1185
(0.0156)
***
-0.0902
(0.0205)
***
BW 2500-3000
-0.0496 ***
(0.0066)
-0.0430 ***
(0.0090)
-0.0581 ***
(0.0120)
-0.0567 ***
(0.0168)
-0.0547
(0.0082)
***
-0.0550
(0.0116)
***
BW 3000-3500
-0.0189 ***
(0.0046)
-0.0243 **
(0.0060)
-0.0167 **
(0.0085)
-0.0218 *
(0.0114)
-0.0192
(0.0057)
***
-0.0220
(0.0078)
***
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
27.74 ***
{40203}
0.18
8.98
***
{40203}
0.69
9.58
***
{12782}
0.19
4.19
***
{12782}
0.74
18.98
{25988}
0.18
***
7.02
{25988}
0.72
***
Gestation<=36 weeks
-0.0843 ***
(0.0115)
-0.0403 ***
(0.0150)
-0.0782 ***
(0.0196)
-0.0533 **
(0.0268)
-0.0894
(0.0140)
***
-0.0510
(0.0192)
***
Gestation 37 weeks
-0.0426 ***
(0.0118)
-0.0345 **
(0.0152)
-0.0564 ***
(0.0209)
-0.0396
(0.0270)
-0.0418
(0.0145)
***
-0.0397
(0.0192)
**
Gestation 38 weeks
-0.0139 *
(0.0072)
-0.0252 ***
(0.0094)
-0.0177
(0.0129)
-0.0285 *
(0.0173)
-0.0158
(0.0088)
*
-0.0304
(0.0121)
**
Gestation 39 weeks
0.0062
(0.0059)
-0.0014
(0.0075)
-0.0105
(0.0106)
-0.0090
(0.0138)
-0.0061
(0.0072)
-0.0089
(0.0096)
Gestation >=42 weeks
-0.0099
(0.0089)
-0.0116
(0.0112)
0.0088
(0.0161)
0.0132
(0.0207)
-0.0035
(0.0109)
-0.0024
(0.0143)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
14.37 ***
{33921}
0.19
3.19
***
{33921}
0.75
4.70
***
{11154}
0.20
1.44
{11154}
0.78
9.53
{22455}
0.19
All regression models include additional fixed effects for mother's martial status, gender of child, and
family sibling size dummies for the birth order of the child within each family size. One, two, and three
asteriks indicate statistical significance at the 10, 5, and 1 percent levels respectively.
-0.0028
(0.0074)
***
2.62
{26204}
0.73
**
0.2182
(0.1535)
***
2.60
{22455}
0.77
**
Table 13
Estimated Effects of Infant Health at Birth on Months on Social Assistance
Close Siblings as Compared to All Siblings, With and Without Family Fixed Effects
Whole Sibling Sample
Sibling Sample
Sibling Sample
No Family F.E.
Family F.E.
Born Within 2 Years
Sibling Sample
Sibling Sample
No Family F.E.
Family F.E.
Born Within 3 Years
Sibling Sample
Sibling Sample
No Family F.E.
Family F.E.
APGAR<=6
1.7872 ***
(0.4518)
0.8932
(0.5716)
-0.0463
(0.8561)
-0.1339
(1.0984)
1.6081
(0.5516)
***
0.7039
(0.7083)
APGAR=7-8
0.6922 ***
(0.1638)
0.2739
(0.2163)
1.6203 ***
(0.3075)
0.5928
(0.4151)
0.9802
(0.1994)
***
0.5445
(0.2676)
APGAR=9
0.2390 **
(0.1005)
-0.2753 **
(0.1375)
0.2405
(0.1885)
-0.5589 **
(0.2650)
0.2284
(0.1216)
*
-0.2471
(0.1705)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
10.32 ***
{40078}
0.07
4.36
***
{40078}
0.58
9.52
***
{12741}
0.07
3.92
***
{12741}
0.64
10.17
{25904}
0.07
***
4.08
{25904}
0.63
BW <1000
-2.3661
(2.2936)
-1.4601
(2.7460)
-3.2997
(5.5915)
-4.0002
(6.9994)
-2.2680
(2.7960)
-1.5733
(3.5186)
BW 1000-1500
0.6699
(0.9757)
0.3227
(1.2513)
-0.1686
(1.6122)
1.6591
(2.0297)
-0.1802
(1.1624)
1.5692
(1.4934)
BW 1500-2500
1.5665 ***
(0.2677)
1.0286 ***
(0.3739)
1.2064 **
(0.4967)
1.0604
(0.7095)
1.6952
(0.3316)
***
1.2814
(0.4699)
***
BW 2500-3000
0.8635 ***
(0.1419)
0.4263 **
(0.2144)
0.7283 ***
(0.2691)
0.3414
(0.4126)
0.9491
(0.1737)
***
0.6068
(0.2665)
**
BW 3000-3500
0.2678 ***
(0.1006)
0.1816
(0.1440)
0.3039
(0.1906)
0.0372
(0.2802)
0.3176
(0.1221)
***
0.1234
(0.1788)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
13.30 ***
{40203}
0.07
1.88 *
{40203}
0.58
2.50
**
{12782}
0.07
0.69
{12782}
0.64
10.25
{25988}
0.07
***
2.23
{25988}
0.63
Gestation<=36 weeks
0.8032 ***
(0.2225)
0.2854
(0.3192)
0.6637
(0.4170)
-0.0647
(0.5797)
1.0075
(0.2748)
***
0.5486
(0.3861)
Gestation 37 weeks
0.4591 **
(0.2286)
0.1549
(0.3233)
0.6457
(0.4445)
0.1013
(0.5841)
0.6541
(0.2841)
**
0.4753
(0.3872)
Gestation 38 weeks
0.1830
(0.1389)
0.0710
(0.2004)
0.2356
(0.2734)
-0.1957
(0.3735)
0.3127
(0.1722)
*
0.1346
(0.2428)
Gestation 39 weeks
-0.1378
(0.1135)
-0.0582
(0.1601)
-0.2032
(0.2260)
-0.3221
(0.2976)
-0.0773
(0.1405)
-0.0044
(0.1937)
Gestation >=42 weeks
0.1828
(0.1728)
-0.0483
(0.2382)
-0.1433
(0.3432)
-0.3928
(0.4477)
0.3221
(0.2142)
-0.1166
(0.2881)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
4.42
***
{33921}
0.07
0.28
{33921}
0.65
1.43
{11154}
0.07
0.39
{11154}
0.74
4.58
{22455}
0.07
All regression models include additional fixed effects for mother's martial status, gender of child, and
family sibling size dummies for the birth order of the child within each family size. One, two, and three
asteriks indicate statistical significance at the 10, 5, and 1 percent levels respectively.
***
0.72
{22455}
0.72
**
***
**
Table 14
Estimated Effects of Infant Health at Birth on Infant Mortality (Death within One Year of Birth)
With and Without Family Fixed Effects
By Parental Residence Income Quintiles
Sibling Sample
No Family F.E.
Family F.E.
1st Income Quintile
No Family F.E.
Family F.E.
1st and 2nd Income Quintile
No Family F.E.
Family F.E.
APGAR Score (Omitted Category APGAR>6)
APGAR<=6
0.3092 ***
(0.0034)
0.3197 ***
(0.0047)
0.2788 ***
(0.0078)
0.2467 ***
(0.0119)
0.3005 ***
(0.0054)
0.3047 ***
(0.0080)
APGAR=7-8
0.0189 ***
(0.0015)
0.0198 ***
(0.0022)
0.0225 ***
(0.0037)
0.0274 ***
(0.0060)
0.0187 ***
(0.0025)
0.0168 ***
(0.0039)
APGAR=9
0.0025 ***
(0.0009)
0.0020
(0.0014)
0.0038 ***
(0.0023)
0.0040
(0.0039)
0.0034 **
(0.0016)
0.0020
(0.0026)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
2887.21 ***
{54091}
0.14
1588.14 ***
{54091}
0.51
444.43 ***
{11958}
0.11
150.62 ***
{11958}
0.60
1073.91 ***
{23043}
0.13
496.82 ***
{23043}
0.59
Birth Weight (Omitted Category BW = 3500+ grams)
BW <1000
0.8572 ***
(0.0071)
0.8723 ***
(0.0099)
0.8225 ***
(0.0168)
0.8025 ***
(0.0270)
0.8733 ***
(0.0124)
0.8872 ***
(0.0181)
BW 1000-1500
0.3622 ***
(0.0066)
0.3912 ***
(0.0093)
0.3570 ***
(0.0162)
0.3638 ***
(0.0248)
0.3876 ***
(0.0110)
0.3978 ***
(0.0164)
BW 1500-2500
0.0479 ***
(0.0022)
0.0630 ***
(0.0034)
0.0410 ***
(0.0048)
0.0493 ***
(0.0084)
0.0448 ***
(0.0035)
0.0627 ***
(0.0058)
BW 2500-3000
0.0065 ***
(0.0012)
0.0130 ***
(0.0021)
0.0086 ***
(0.0029)
0.0082
(0.0056)
0.0082 ***
(0.0020)
0.0131 ***
(0.0037)
BW 3000-3500
0.0018 **
(0.0009)
0.0042 ***
(0.0014)
0.0034
(0.0022)
0.0056
(0.0039)
0.0037 **
(0.0015)
0.0065 ***
(0.0026)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
3586.36 ***
{54310}
0.25
1899.91 ***
{54310}
0.57
584.69 ***
{12018}
0.21
217.44 ***
{12018}
0.63
1265.61 ***
{23157}
0.22
590.8
{23157}
0.62
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
0.1043 ***
(0.0023)
0.1187 ***
(0.0039)
0.0929 ***
(0.0051)
0.1078 ***
(0.0095)
0.1012 ***
(0.0036)
0.1216 ***
(0.0066)
Gestation 37 weeks
0.0077 ***
(0.0025)
0.0142 ***
(0.0041)
0.0114 **
(0.0055)
0.0111
(0.0101)
0.0069 *
(0.0040)
0.0120 *
(0.0071)
Gestation 38 weeks
0.0050 ***
(0.0015)
0.0073 ***
(0.0026)
0.0047
(0.0037)
0.0080
(0.0068)
0.0036
(0.0026)
0.0108 **
(0.0046)
Gestation 39 weeks
0.0009
(0.0013)
-0.0003
(0.0021)
0.0002
(0.0032)
-0.0043
(0.0058)
-0.0009
(0.0022)
-0.0014
(0.0039)
Gestation >=42 weeks
0.0002
(0.0019)
0.0005
(0.0031)
-0.0009
(0.0049)
-0.0051
(0.0092)
-0.0015
(0.0033)
-0.0028
(0.0058)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
435.15 ***
{45583}
0.05
198.70 ***
{45583}
0.51
162.82 ***
{19228}
0.05
73.48
{19228}
0.58
70.68
{9914}
0.05
***
28.44
{9914}
0.59
***
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies for the birth order of the
child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and 1 percent levels respectively.
Table 15
Estimated Effects of Infant Health at Birth on Grade 12 Attainment by Age 17
With and Without Family Fixed Effects
By Parental Residence Income Quintiles
Sibling Sample
No Family F.E.
Family F.E.
1st Income Quintile
No Family F.E.
Family F.E.
1st and 2nd Income Quintile
No Family F.E.
Family F.E.
APGAR Score (Omitted Category APGAR>6)
APGAR<=6
-0.0739 ***
(0.0206)
-0.0410 *
(0.0239)
-0.0897 *
(0.0526)
-0.0069
(0.0621)
-0.0870 **
(0.0364)
0.0028
(0.0426)
APGAR=7-8
-0.0302 ***
(0.0075)
-0.0203 **
(0.0091)
-0.0879 ***
(0.0208)
-0.0216
(0.0269)
-0.0656 ***
(0.0137)
-0.0241
(0.0170)
APGAR=9
0.0111 **
(0.0046)
0.0051
(0.0058)
-0.0338 ***
(0.0125)
0.0069
(0.0168)
-0.0179 **
(0.0084)
0.0194 *
(0.0110)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
16.3
***
{40514}
0.18
4.12
***
{40514}
0.70
0.47
{6482}
0.73
9.02
***
{14109}
0.19
3.09
**
{14109}
0.72
6.86
{6482}
0.18
***
Birth Weight (Omitted Category BW = 3500+ grams)
BW <1000
-0.1308
(0.1060)
0.0546
(0.1153)
-0.5009
(0.3186)
-0.1747
(0.3469)
-0.2229
(0.2018)
-0.0754
(0.2103)
BW 1000-1500
-0.1271 ***
(0.0451)
-0.1410 ***
(0.0525)
0.0223
(0.1097)
-0.1117
(0.1275)
-0.0398
(0.0813)
-0.1608 *
(0.0938)
BW 1500-2500
-0.1168 ***
(0.0124)
-0.0815 ***
(0.0157)
-0.0724 **
(0.0309)
-0.1175 ***
(0.0420)
-0.1076 ***
(0.0207)
-0.1206 ***
(0.0272)
BW 2500-3000
-0.0496 ***
(0.0066)
-0.0430 ***
(0.0090)
-0.0439 **
(0.0172)
-0.0814 ***
(0.0245)
-0.0414 ***
(0.0119)
-0.0596 ***
(0.0167)
BW 3000-3500
-0.0189 ***
(0.0046)
-0.0243 **
(0.0060)
-0.0324 **
(0.0128)
-0.0672 ***
(0.0173)
-0.0187 **
(0.0086)
-0.0466 ***
(0.0115)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
27.74 ***
{40203}
0.18
8.98
***
{40203}
0.69
7.24
***
{13975}
0.19
6.12
***
{13975}
0.72
3.07
{6402}
0.18
***
4.02
{6402}
0.73
***
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
-0.0843 ***
(0.0115)
-0.0403 ***
(0.0150)
-0.0642 **
(0.0267)
-0.0334
(0.0370)
-0.0839 ***
(0.0194)
-0.0343
(0.0257)
Gestation 37 weeks
-0.0426 ***
(0.0118)
-0.0345 **
(0.0152)
-0.0621 **
(0.0300)
-0.0768 *
(0.0406)
-0.0490 **
(0.0213)
-0.0561 **
(0.0276)
Gestation 38 weeks
-0.0139 *
(0.0072)
-0.0252 ***
(0.0094)
-0.0265
(0.0190)
-0.0466 *
(0.0258)
-0.0259 **
(0.0131)
-0.0247
(0.0174)
Gestation 39 weeks
0.0062
(0.0059)
-0.0014
(0.0075)
-0.0219
(0.0164)
-0.0463 **
(0.0221)
-0.0129
(0.0111)
-0.0258 *
(0.0146)
Gestation >=42 weeks
-0.0099
(0.0089)
-0.0116
(0.0112)
0.0106
(0.0245)
-0.0074
(0.0317)
-0.0134
(0.0166)
-0.0082
(0.0216)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
14.37 ***
{33921}
0.19
3.19
***
{33921}
0.75
4.80
***
{11829}
0.19
1.41
{11829}
0.76
2.25
{5426}
0.18
**
1.53
{5426}
0.77
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies for the birth order of the
child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and 1 percent levels respectively.
Table 16
Estimated Effects of Infant Health at Birth on Social Assistance Take-up (By months)
With and Without Family Fixed Effects
By Parental Residence Income Quintiles
Sibling Sample
No Family F.E.
Family F.E.
1st Income Quintile
No Family F.E.
Family F.E.
1st and 2nd Income Quintile
No Family F.E.
Family F.E.
APGAR Score (Omitted Category APGAR>6)
APGAR<=6
1.7872 ***
(0.4518)
0.8932
(0.5716)
0.6678
(1.5623)
-1.6881
(1.8615)
0.6372
(0.9222)
-1.2344
(1.1105)
APGAR=7-8
0.6922 ***
(0.1638)
0.2739
(0.2163)
1.0803 *
(0.6131)
1.0298
(0.8087)
0.6451 *
(0.3434)
0.6462
(0.4450)
APGAR=9
0.2390 **
(0.1005)
-0.2753 **
(0.1375)
0.5193
(0.3705)
-0.0592
(0.5034)
0.6100 ***
(0.2100)
-0.3262
(0.2868)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
10.32 ***
{40078}
0.07
4.36 ***
{40078}
0.58
1.25
{6373}
0.14
1.08
{6373}
0.69
3.09
**
{13911}
0.10
2.41
*
{13911}
0.66
Birth Weight (Omitted Category BW = 3500+ grams)
BW <1000
-2.3661
(2.2936)
-1.4601
(2.7460)
-3.1673
(9.3392)
-2.0083
(10.4347)
-1.3138
(5.0493)
-0.0190
(5.5022)
BW 1000-1500
0.6699
(0.9757)
0.3227
(1.2513)
-3.4778
(3.2147)
-0.2180
(3.8347)
-1.9954
(2.0333)
-1.0106
(2.4544)
BW 1500-2500
1.5665 ***
(0.2677)
1.0286 ***
(0.3739)
0.6897
(0.9044)
0.1894
(1.2630)
0.8926 *
(0.5188)
0.3344
(0.7125)
BW 2500-3000
0.8635 ***
(0.1419)
0.4263 **
(0.2144)
0.6625
(0.5044)
0.0412
(0.7375)
0.8007 ***
(0.2976)
-0.2521
(0.4378)
BW 3000-3500
0.2678 ***
(0.1006)
0.1816
(0.1440)
0.4772
(0.3742)
0.9332 *
(0.5212)
0.4010 *
(0.2152)
0.3290
(0.3000)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
13.03 ***
{40203}
0.07
1.88
*
{40202}
0.58
2.25
**
{13975}
0.10
0.62
{13975}
0.66
0.85
{6402}
0.14
0.83
{6402}
0.69
Gestation (Omitted Category 40-41 Weeks)
Gestation<=36 weeks
0.8032 ***
(0.2225)
0.2854
(0.3192)
1.0729
(0.7058)
0.2518
(0.9722)
0.9188 **
(0.4384)
0.4505
(0.6044)
Gestation 37 weeks
0.4591 **
(0.2286)
0.1549
(0.3233)
0.3120
(0.7919)
-0.4744
(1.0656)
0.4372
(0.4808)
-0.2881
(0.6484)
Gestation 38 weeks
0.1830
(0.1389)
0.0710
(0.2004)
0.6034
(0.5020)
-0.4045
(0.6776)
0.5134 *
(0.2975)
-0.3208
(0.4096)
Gestation 39 weeks
-0.1378
(0.1135)
-0.0582
(0.1601)
0.1151
(0.4340)
-0.2642
(0.5819)
0.0341
(0.2512)
-0.0645
(0.3439)
Gestation >=42 weeks
0.1828
(0.1728)
-0.0483
(0.2382)
0.5295
(0.6469)
-0.2873
(0.8325)
0.5239
(0.3748)
-0.2958
(0.5071)
F-test:No Inf. Hlth. Effects
Sample Size
R-Squared
4.42 ***
{33921}
0.07
0.28
{33921}
0.65
0.73
{5426}
0.14
0.17
{5426}
0.76
1.66
{11829}
0.10
0.39
{11829}
0.72
All regression models include additional fixed effects for mother's martial status, gender of child, and family sibling size dummies for the birth order of the
child within each family size. One, two, and three asteriks indicate statistical significance at the 10, 5, and 1 percent levels respectively.
Table A1
Estimated Effects of Infant Health at Birth on Infant Mortality (Death within One Year of Birth)
With and Without Family Fixed Effects
Linear Model
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
With Family F.E.
Twins Sample
No Family F.E.
Twins Sample
With Family F.E.
-0.055
(0.003)
-0.013
(0.004)
APGAR Score
Coefficient Estimate
Standard Error
-0.036
(0.000)
***
-0.041
(0.000)
***
-0.044
(0.001)
***
***
***
Birth Weight (in grams)
Coefficient Estimate
Standard Error
-3.8E-05 ***
5.1E-07
-4.4E-05 ***
8E-07
-7.1E-05 ***
1.46E-06
-9.3E-05 ***
5.11E-06
-1.6E-05
1.2E-05
Gestation (in weeks)
Coefficient Estimate
Standard Error
-0.015
(0.000)
***
-0.017
(0.000)
***
-0.022
(0.000)
***
-0.024
(0.001)
***
NA
NA
All regression models include additional fixed effects for mother's martial status, mother's age at child's birth, gender of child,
and family sibling size dummies for the birth order of the child within each family size. One, two, and three atericies indicate
statistical significance at the 10, 5, and 1 percent levels respectively.
Table A2
Estimated Effects of Infant Health at Birth on Physician Visits
With and Without Family Fixed Effects
Linear Model
Mean Age:
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
With Family F.E.
Twins Sample
No Family F.E.
Twins Sample
With Family F.E.
APGAR Score
Coefficient Estimate
Standard Error
-0.3091 ***
(0.0569)
-0.2164 ***
(0.0793)
0.0459
(0.0912)
-0.1233
(0.2705)
-0.1306
(0.3491)
Birth Weight (in grams)
Coefficient Estimate
Standard Error
-1.62E-04 *
8.46E-05
-5.32E-05
1.19E-04
-8.26E-05
1.77E-04
-1.27E-03 *
5.27E-04
-1.86E-04
9.94E-04
Gestation (in weeks)
Coefficient Estimate
Standard Error
-0.0515 *
(0.0284)
0.0073
(0.0408)
0.0486
(0.0548)
-0.0952
(0.1120)
NA
NA
All regression models include additional fixed effects for mother's martial status, mother's age at child's birth, gender of child,
and family sibling size dummies for the birth order of the child within each family size. One, two, and three atericies indicate
statistical significance at the 10, 5, and 1 percent levels respectively.
Table A3
Estimated Effects of Infant Health at Birth on Language Arts Score
With and Without Family Fixed Effects
Linear Model
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
-0.018
(0.023)
-0.029
(0.025)
9.08E-05 **
4.44E-05
-1.8E-05
7.06E-05
APGAR Score
Coefficient Estimate
Standard Error
0.021
(0.004)
***
0.018
(0.006)
***
0.011
(0.006)
*
Birth Weight (in grams)
Coefficient Estimate
Standard Error
8.1E-05 ***
0.000006
9.9E-05 ***
8.76E-06
3.47E-05 ***
1.26E-05
Gestation (in weeks)
Coefficient Estimate
Standard Error
0.009
(0.002)
***
0.013
(0.003)
***
-0.004
(0.004)
0.006
(0.010)
NA
NA
All regression models include additional fixed effects for mother's martial status, mother's age at child's birth, gender of child,
and family sibling size dummies for the birth order of the child within each family size. One, two, and three atericies indicate
statistical significance at the 10, 5, and 1 percent levels respectively.
A4
Estimated Effects of Infant Health at Birth on Attaining Grade 12 by Age 17
With and Without Family Fixed Effects
Linear Model
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
-0.001
(0.011)
-0.005
(0.011)
7.41E-05 ***
2.14E-05
2.71E-05
3.19E-05
APGAR Score
Coefficient Estimate
Standard Error
0.010
(0.002)
***
0.010
(0.003)
***
0.007
(0.003)
**
Birth Weight (in grams)
Coefficient Estimate
Standard Error
3.46E-05 ***
2.76E-06
4.48E-05 ***
4.02E-06
4.3E-05 ***
6.29E-06
Gestation (in weeks)
Coefficient Estimate
Standard Error
0.007
(0.001)
***
0.009
(0.001)
***
0.006
(0.002)
***
0.008
(0.005)
*
NA
NA
All regression models include additional fixed effects for mother's martial status, mother's age at child's birth, gender of child,
and family sibling size dummies for the birth order of the child within each family size. One, two, and three atericies indicate
statistical significance at the 10, 5, and 1 percent levels respectively.
Table A5
Estimated Effects of Infant Health at Birth on Graduation on Time
With and Without Family Fixed Effects (1982 and later births only)
Linear Model
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
0.010
(0.019)
-0.017
(0.020)
3.7E-06
3.46E-05
2.48E-05
5.39E-05
0.011
(0.007)
NA
NA
APGAR Score
Coefficient Estimate
Standard Error
0.010
(0.003)
***
0.016
(0.006)
***
0.013
(0.008)
Birth Weight (in grams)
Coefficient Estimate
Standard Error
1.96E-05 ***
4.45E-06
3.79E-05 ***
9.33E-06
3.36E-05 **
1.54E-05
Gestation (in weeks)
Coefficient Estimate
Standard Error
0.002
(0.001)
0.006
(0.003)
**
-0.005
(0.004)
All regression models include additional fixed effects for mother's martial status, mother's age at child's birth, gender of child,
and family sibling size dummies for the birth order of the child within each family size. One, two, and three atericies indicate
statistical significance at the 10, 5, and 1 percent levels respectively.
Table A6
Estimated Effects of Infant Health at Birth on Social Assistance (yes/no) After Age 18
With and Without Family Fixed Effects
Linear Model
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
-0.001
(0.007)
-0.012
(0.010)
-3.8E-05 ***
1.3E-05
-3.5E-05
2.73E-05
APGAR Score
Coefficient Estimate
Standard Error
-0.010
(0.001)
***
-0.010
(0.002)
***
-0.001
(0.002)
Birth Weight (in grams)
Coefficient Estimate
Standard Error
-1.7E-05 ***
1.8E-06
-1.7E-05 ***
2.62E-06
-5.3E-06
4.23E-06
Gestation (in weeks)
Coefficient Estimate
Standard Error
-0.003
(0.001)
***
-0.003
(0.001)
**
-0.001
(0.001)
-0.004
(0.003)
NA
NA
All regression models include additional fixed effects for mother's martial status, mother's age at child's birth, gender of child,
and family sibling size dummies for the birth order of the child within each family size. One, two, and three atericies indicate
statistical significance at the 10, 5, and 1 percent levels respectively.
Table A7
Estimated Effects of Infant Health at Birth on Social Assistance (months) After Age 18
With and Without Family Fixed Effects
Linear Model
Full Sample
Sibling Sample
No Family F.E.
Sibling Sample
Family F.E.
Twins Sample
No Family F.E.
Twins Sample
Family F.E.
-0.188
(0.210)
-0.273
(0.322)
-0.00017
0.00041
-0.00081
0.000914
-0.099
(0.068)
NA
NA
APGAR Score
Coefficient Estimate
Standard Error
-0.334
(0.040)
***
-0.305
(0.058)
***
-0.078
(0.077)
Birth Weight (in grams)
Coefficient Estimate
Standard Error
-0.00057 ***
5.91E-05
-0.00061 ***
8.71E-05
-0.00039 ***
0.00015
Gestation (in weeks)
Coefficient Estimate
Standard Error
-0.079
(0.018)
***
-0.075
(0.027)
***
-0.051
(0.041)
All regression models include additional fixed effects for mother's martial status, mother's age at child's birth, gender of child,
and family sibling size dummies for the birth order of the child within each family size. One, two, and three atericies indicate
statistical significance at the 10, 5, and 1 percent levels respectively.